openfast icon indicating copy to clipboard operation
openfast copied to clipboard
verified publisher icon OpenFAST

Bug report: AeroDyn Write Output Not Appearing in Linearization

Open jjonkman opened this issue 3 years ago • 1 comments

Bug description When write outputs are specified in AeroDyn for an OpenFAST model that is linearized, the derivatives of these outputs should be calculated and reported in the linearized C and D matrices. However, the rows in the C and D matrices associated with the AeroDyn write outputs are all zero. This functionality worked in OpenFAST v2.6 and older, but no longer works in OpenFAST v3.0 and the current dev branch. I would guess the bug originated when the multi-rotor capability was added to AeroDyn in PR # https://github.com/OpenFAST/openfast/pull/672.

To Reproduce See the model input files shared by Dezheng Zhu on the forum: https://wind.nrel.gov/forum/wind/viewtopic.php?f=30&t=2667 and compare the C and D matrices between OpenFAST v2.6 and v3.0.

Expected behavior Nonzero derivatives associated with the AeroDyn write outputs.

Screenshots, if applicable

OpenFAST Version OpenFAST v3.0 and current dev.

System Information (please complete the following information): Any

Additional context

jjonkman avatar Jan 19 '22 18:01 jjonkman

We might have to

  1. call AD_CalcOutput instead of RotCalcOutput in Rot_JacobianPInput() and RotJacobianPContState() or
  2. move where y%rotors(iR)%WriteOutput is computed from AD_CalcOutput into RotCalcOutput.

bjonkman avatar Jan 19 '22 18:01 bjonkman

This issue was resolved by PR #1061.

jjonkman avatar Feb 10 '23 15:02 jjonkman

Reopening this issue as we now thing we should be calling AD_CalcOutput.

This became an issue after the addition of the buoyancy loads to AD15. These loads were added to the AD_CalcOutput, so would not get included in the linearization results. So we think it best to not have the glue code aware of the rotors in AD15.

andrew-platt avatar Jun 05 '23 22:06 andrew-platt

I don't think this is a matter of the glue code being aware of rotors in AD15.... For instance, the glue code calls AD_JacobianPInput, which calls Rot_JacobianPInput. I don't think the rotor-level Jacobian routines should be calling the AD (all rotors+) routines, so you'd have to do some other rearranging of those jacobian routines (not just calling AD_CalcOutput).

bjonkman avatar Jun 06 '23 04:06 bjonkman